[Biojava-l] Java Resource Management [a semi troll...]

[Thomas Down]

On Sun, Feb 09, 2003 at 09:24:07PM +0000, Ewan Birney wrote:
> 
> 
> I know... this is a troll.... but it sort of resonates with me about the
> way Java development works - people have to fix on a JVM version to get
> real like-a-rock stability and work about bugs in that release, and 
> Java is just dreadful in resource management (in particular memory
> footprint) meaning you have to give it really beefy machines to run on.
> 
> 
> 
> http://www.internalmemos.com/memos/memodetails.php?memo_id=1321

There seem to be a number of points mixed up in the document,
but the one which they were emphasizing most was memory usage.
That's something which I've found to be very VM-dependant (I guess
not surprisingly).

I make fairly heavy use of the jEdit text editor
(http://jedit.sf.net/, for anyone who's interested).  Mainly
I run this on Sun JDK1.4.1, Client VM, Linux-intel, and it
works pretty well.  Even with a good collection (10-15) source
files open, it's generally only using 4Mb of Java heap space
(i.e. actual Java objects).  At this point, the RSS of the
java process is around 40Mb.  But the bulk of the overhead is
in shared libraries (>20Mb).  The overhead for applications which
don't use the windowing toolkit is *much* lower (just a few
megabytes).

In contrast, jEdit running on the Compaq/HP/whoever fast
VM on an alpha uses around 200Mb of RSS, with the Java heap
hitting >50Mb.  Even being generous and allowing a factor
of two for all those 64 bit pointers, that's a pretty dramatic
difference...

Anyway, the difference between two VMs (both fairly modern
mixed-mode designs) seems to be more significant than any comparison
I've yet seen between Java and Language X.

In spite of this, I don't want to know what TogetherJ is doing
under the hood...

    Thomas.